Article pubs.acs.org/Langmuir
Analysis Method for Quantifying the Morphology of Nanotube Networks Dusan Vobornik,* Shan Zou, and Gregory P. Lopinski Measurement Science and Standards, National Research Council Canada, 100 Sussex Drive, Ottawa, Ontario K1A 0R6, Canada S Supporting Information *
ABSTRACT: While atomic force microscopy (AFM) is a powerful technique for imaging assemblies and networks of nanoscale materials, approaches for quantitative assessment of the morphology of these materials are lacking. Here we present a volume-based approach for analyzing AFM images of assemblies of nano-objects that enables the extraction of relevant parameters describing their morphology. Random networks of single-walled carbon nanotubes (SWCNTs) deposited via solution-phase processing are used as an example to develop the method and demonstrate its utility. AFM imaging shows that the morphology of these networks depends on details of processing and is influenced by choice of substrate, substrate cleaning method, and postdeposition rinsing protocols. A method is outlined to analyze these images and extract relevant parameters describing the network morphology such as the density of SWCNTs and the degree to which tubes are bundled. Because this volume-based approach depends on accurate measurements of the height of individual tubes and their networks, a procedure for obtaining reliable height measurements is also discussed. Obtaining quantitative parameters that describe the network morphology allows going beyond qualitative descriptions of images and will facilitate optimizing network preparation methods based on measurable criteria and correlating performance with morphology.
■
INTRODUCTION Nano-objects such as nanotubes, nanowires, nanosheets, and nanoparticles continue to be of interest as building blocks for functional materials due to their remarkable size-dependent properties. However, the properties of materials constructed from these objects depend not only on the properties of the objects themselves but also on how these blocks assemble into larger structures.1−3 Although electron and scanned probe microscopies are commonly used to visualize the morphology of these assemblies, methods for quantitative assessment of the resulting images have received less attention. Extraction of quantitative parameters describing the morphology of a sample from images will facilitate feedback on how processing affects the structure and ultimately how the structure influences the properties of the material. While the analysis approach presented here should be widely applicable to a range of nanoscale materials, random networks of single-walled carbon nanotubes deposited from solution are used as an example to illustrate the method and demonstrate its utility. These networks represent interesting model systems for investigating the interplay between the intrinsic properties of the individual nanoscale building blocks and process-dependent network morphologies in determining properties. While individual single-walled carbon nanotubes (SWCNTs) exhibit high intrinsic conductivities and field effect mobilities, films based on random networks of these tubes show considerably lower values.1,4−7 Over distances greater than the length of an individual tube, electrical transport is usually limited by tube− tube junctions, making the conductivity highly dependent on Published 2016 by the American Chemical Society
details of the network morphology (i.e., tube density, bundling, and alignment).8−12 Furthermore, starting with the same carbon nanotube ink, process details can strongly influence the morphology and consequently the electronic properties of the network.13 Atomic force microscopy (AFM) is a powerful and versatile probe of nanomaterial morphologies, enabling imaging of the individual nano-objects and their assemblies in a variety of environments (vacuum, ambient, liquid) and regardless of whether these materials are insulating or conducting. Quantitative AFM studies of nanomaterials have typically focused on extracting distributions of lengths, heights, or diameters for individual nano-objects.14−19 Obtaining this type of data usually requires optimizing sample preparation conditions so that isolated features can be imaged on a flat substrate. However, functional assemblies are most often achieved at higher densities. For example, the formation of conductive networks of nanowires and nanotubes for applications such as transparent conductive electrodes or channel materials for thin film transistors (TFTs) requires densities above the percolation threshold. At these higher densities there is likely to be some degree of overlap and aggregation of the individual building blocks. In these realistic applications it is often hard to see where one nano-object ends and the other starts, making it difficult to accurately count the Received: July 4, 2016 Revised: August 5, 2016 Published: August 10, 2016 8735
DOI: 10.1021/acs.langmuir.6b02475 Langmuir 2016, 32, 8735−8742
Langmuir
Article
different ways are shown in Figure 1. The rather different morphologies readily apparent in these images illustrate how
exact number of objects in the network. Among several measurands that can be used to characterize the morphology of a random network of nanowires or tubes, two that are particularly important in determining electronic and optical properties are the tube density and the extent to which the individual tubes aggregate into bundles. In this work a straightforward and fast volume-based method to extract these measurands from experimentally obtained AFM images is presented. The use of a volume-based analysis method means that the accuracy of AFM height measurements (from which the volume is calculated) is of paramount importance. Therefore, we also propose an experimental procedure that facilitates verification of the AFM imaging parameters to ensure reliable measurements of the nanotube height.
■
EXPERIMENTAL SECTION
Substrate Preparation. Two different substrates were used; a thermally grown SiO2 thin film on silicon, and highly ordered pyrolytic graphite (HOPG). A silicon wafer with a 100 nm thick thermal oxide (Silicon Quest International) was cut into 1 cm2 pieces. Prior to nanotube network deposition the silicon oxide surface was cleaned by either: (a) Piranha solution bath (3:1 volume ratio of 98% H2SO4 and 30% H2O2) for 30 min, followed by thorough rinsing with ultrapure water (resistivity of 18.2 MΩ·cm), and blown dry with nitrogen; or (b) 5 min in an oxygen plasma cleaner (Yield Engineering Systems G500). For the HOPG substrates, we used ZYB-grade, 12 mm × 12 mm HOPG squares (Bruker AFM Probes, Camarillo, CA). Clean surfaces were obtained by cleaving off the top layers with Scotch tape prior to nanotube network deposition. SWCNT Network Preparation. A commercially available ultrahigh purity semiconducting SWCNT dispersion was purchased from Nanointegris (http://www.nanointegris.com/IsoSol-S100). The separation and purity of these nanotubes are ensured by the poly(9,9-di-ndodecylfluorene) (PFDD) wrapping.20 Networks were prepared by dropcasting 40 μL of a 10 mg/L toluene solution of the nanotubes on clean substrates and letting the toluene evaporate. This typically took 10 to 15 min. To remove excess polymer (initial polymer to nanotube mass ratio was 4 to 1), as well as any other contaminants, we rinsed the samples upon solvent evaporation with a steady stream of toluene for 20 s or by successive 20 s rinses of toluene, tetrahydrofuran (THF), and isopropanol (IPA). Finally, samples were dried with nitrogen and stored in a closed Petri dish under room conditions. AFM measurements were carried within a day of preparation, but several samples were again measured at different times during several months following their preparation with no significant changes in morphology observed. Chemicals. Toluene and tetrahydrofuran were purchased from EMD Millipore with respective purities (GC) of ≥99.5 and ≥99.9%. Distilled in glass-grade isopropanol was purchased from Caledon Chemicals with a purity (GC) of ≥99.7%. AFM Imaging. The samples were imaged using the MultiMode AFM with the NanoScope V controller (Bruker Nano Surfaces Division, Santa Barbara, CA) in Bruker’s proprietary PeakForce QNM mode. The peak force with which the tip taps the sample surface was always kept close to the lowest stable imaging level of 0.5 nN or less (stable here means perfectly overlapped trace and retrace lines during AFM scanning). We have used ScanAsyst-Air AFM probes (Bruker AFM Probes, Camarillo, CA), which are made of silicon nitride and whose typical tip radius is 2 nm according to the manufacturer’s specifications. Analysis Software. All analysis of AFM images was performed using Gwyddion, a free, open-source software, with well-defined and explained operations and functions.21
Figure 1. AFM images of networks obtained by dropcasting the same solution of carbon nanotubes on SiO2 (a−c) and HOPG (d). Prior to deposition, SiO2 substrates were cleaned either by Piranha solution (a,b) or by oxygen plasma treatment (c), while HOPG was freshly cleaved (d). Upon solvent evaporation samples were rinsed with toluene for 20 s (a,c,d) or sequentially with toluene, tetrahydrofuran, and isopropanol for 20 s each (b). All images have the same 1 μm2 size and are displayed with the same 9 nm vertical scale, where 0 corresponds to the lowest pixel height in the image. Cross sections in panel e correspond to numbered lines shown in panel a.
details of sample processing influence network formation, even starting from the same SWCNT dispersion. Specifically, the observed network variations result from the use of different substrates, different substrate cleaning procedures, or different postdeposition rinsing procedures, as detailed in the figure caption. It is easy to qualitatively observe certain differences between the networks in Figure 1. For example, there seems to be more tubes on the oxygen plasma-cleaned (Figure 1c) versus piranha-cleaned SiO2 surface (Figure 1a). Similarly, it appears that additional rinsing with tetrahydrofuran and isopropanol (Figure 1b) leads to more features greater than 5 nm in height, indicative of substantial aggregation (bundling) of the SWCNTs, yet putting numbers on these differences appears to be very difficult. Some representative cross sections from Figure 1a (numbered white lines) are shown in Figure 1e. While the SWCNTs used to make the dispersions used here have a
■
RESULTS AND DISCUSSION Process Details Influence SWCNT Network Morphology. AFM images of SWCNT networks processed in slightly 8736
DOI: 10.1021/acs.langmuir.6b02475 Langmuir 2016, 32, 8735−8742
Langmuir
Article
Figure 2. (a) Scheme showing a bundle of seven nanotubes. The bundle’s projected AFM surface is shown in red, and together with the green surface on the top it illustrates the boundaries within which the volume of the bundle is calculated. (b) Cross section showing a 1.7 nm diameter nanotube being scanned by a 5 nm diameter AFM probe. The red line corresponds to the trajectory of the lowest point of the tip during scanning. The green line shows a similar trajectory that would result from a 10 nm diameter AFM tip scanning the same nanotube, and the orange line shows the ideally deconvoluted trajectory. (c) Trajectories for the 5 and 10 nm diameter AFM tips are shown in dashed red and green lines, the ideal deconvoluted contour is shown in orange, and the trapezoid shaped approximated contours that we used in calculations carried out in panel d are shown in solid red and green lines. (d) Table based on dividing the calculated AFM volume of a bundle by the calculated AFM volume of a single nanotube for three different AFM tip diameters as well as for deconvoluted AFM images. The table highlights the error of the volume based analysis if the deconvolution is not applied.
these images. These parameters can then be used to compare different network fabrication process and correlate resulting morphologies with properties of the networks. Volumetric AFM Analysis for Carbon Nanotube Networks. Volumetric analysis of AFM data has been proposed in the past, but it was based on an apparent mass− volume relationship15 and did not offer adequate solutions for high surface density, partly overlapping, or bundled samples. In this work, we present a volumetric analysis method that does not rely on any mass−volume equivalency and which works well even for dense and highly bundled samples. Using carbon nanotube networks, we show how this volume-based analysis can quantify the degree of bundling and offer a straightforward analysis method for a dense network, where it is often impossible to make out individual nanotubes. As discussed above (and also highlighted in the part 1 of the Supporting Information), a particular challenge when trying to quantify nanotube assemblies is the tendency for individual tubes to bundle. Even with the use of polymers to disperse the nanotubes, most of the SWCNTs in a realistic network are observed to be bundled. The method proposed here is based on a simple hypothesis that the volume of a bundle of nanotubes is equal to the product of the volume of a single nanotube by the number of nanotubes in the bundle
narrow distribution of diameters ranging from 1.2 to 1.4 nm, simple geometric analysis shows that if there was no polymer wrapped and we took the three smallest 1.2 nm diameter nanotubes and bundled them together into a tight pack as illustrated in the Supporting Information Figure S1 (i), the bundle height would still not exceed 2.24 nm. On the basis of this, network features whose height is above 2 nm are considered to be bundles consisting of several nanotubes with some degree of vertical stacking, while features whose height is
